Problem with Running xtlsdvc command

[Love Idahosa]

Hello,

I am currently trying to run the xtlsdvc command on the STATA 12.1 but keep encountering issues.
The problem is that when I run the command on the full data set, STATA hangs/freezes indefinitely.
M computer is a runs on a 64bit Windows 7 Professional Operating System with 4GB RAM.
I have tried running the code on a different computer with 8GB RAM but still get the same issue.
I have detailed the problem as well as the solutions I have tried and the outcomes below.
I would like to know if the freezing/hanging is a syntax problem, a data problem, an operating system problem, or something else.

The details of the problem are as follows:

My dataset is an unbalanced panel with 21 cross section units, 21,550 observations, and no gaps.
I run the following xtlsdvc command based on the following syntax in the paper titled "Estimation and inference in dynamic unbalanced panel-data models with a small number of individuals" by Giovanni S.F, Bruno as detailed in "The Stata Journal (2005) 5, Number 4, pp. 473-500"

SYNTAX: xtlsdvc depvar [indepvars] [if] , initial(estimator)
-See page 497

The model I estimate is:

xtlsdvc consumptionkwh temperature humidity2, i(ah)
- I also run this model by replacing the consistent estimator chosen to initialize the bias correction with i(ab) and i(bb) respectively.

Problem: In all cases, when I run the command on the full data set, Stata hangs/Freezes.

I have tried out the following alternatives to running the code and had the following results:

1. Reduce the sample size by restricting to a small category (i.e. I restricted the data to run for a specific year since my data frequency is daily).
Results:
- For the i(ah) option; when I reduce the sample size by restricting the sample to a certain category, the command works.
- Reducing the sample size for the i(ab) option return a r(2000) error saying no observations.
- For the i(bb) option however, I get a r(3900) error which I understand is a problem with size of my Operating system.

2. Condense the frequency of the data from daily observations(i.e. 365 per year) to monthly observations (i.e. 12 per year) by aggregating or averaging the figures.
Results:
-In all option cases [i.e. i(ah), i(ab), and i(bb)], I am able to get the codes to run for the entire data set without having to restrict the sample to a small category.

I would like to know if the freezing/hanging is a syntax problem, a data problem, an operating system problem, or something else; and what I can do to solve the problem.
.

Thank you.

[Friedrich Huebler]

Welcome to Statalist and thank you for the detailed description of your problem.

It is helpful for other list members if you show the exact commands that you used and the exact output returned by Stata. Please do not retype but copy and paste directly from the Results window of Stata. Please post the commands and Stata output with CODE tags. You can find more advice about asking questions on Statalist in the FAQ, especially section 12.

[Sebastian Kripfganz]

Coincidentally, I ran into a similar problem with xtlsdvc today. It indeed seems that the command hangs / freezes when the data set is very large. Yet, I have not investigated the issue further. Maybe I just was not patient enough.

In your case, however, I wonder why you want to use this command in the first place. With only 21 cross-sectional units and many time periods you are clearly in a large-T situation. xtlsdvc corrects for a small-T bias in dynamic panel models. This should not be something you need to worry about.

[Daniel Weimar]

Hi. I am also struggling with the "freeze" problem, although I use a small panel with 1343 observations (79 (N) x 17(T)). Perhaps, the problem is associated with the Stata version (14)?!